Search

Recent European and German climate targets call for a faster power system transition towards variable renewable energy sources. With the increasing importance of Open Science, several Open Source models have been developed in recent years. However, only a few studies exist that compare their performance against each other. Therefore, this study performs a comprehensive model comparison of five mature Open Source power sector models. For this purpose, we apply eight fully harmonized and simplified one-year scenarios for the German power sector, to analyze deviations in model results. First, an in-depth analysis of two base scenarios for 2016 and 2030 reveals that linear programming-based models differ substantially from models with pre-implemented dispatch orders. Other deviations occur across all models and are mainly caused by the indifferent use of flexibility options such as storage and transmission. Second, variations of parameters and characteristics with a political significance are individually applied to the 2030 base scenario to identify their impact on model results. This includes CO 2 emission budgets, increased demands by sector coupling, coal exit strategies, and renewable generation shares. The results prove that some models are far more sensitive to these parameters than others, and renewable generation shares alone are not sufficient to reach desired effects in emission reductions. Finally, a comprehensive scenario for 2030 combines all measures to evaluate general trends that result from the most recent updates in German energy policy. Model results indicate that the new targets require substantially increased investments into renewable generation capacities, storage, and transmission. • Comprehensive scenario-based model comparison with five mature Open Source power system models. • Evaluation of eight scenarios that model the transition of the German power sector until 2030. • Linear programming-based models differ substantially from models with an implemented dispatch structure. • Sensitivity towards policy-relevant modeling parameters is more or less pronounced, depending on the used model. • Updated political targets call for substantially increased investments into renewable and flexible technologies. [ABSTRACT FROM AUTHOR]

Energy system modeling is a commonly used method to provide policy recommendations and insight to transformation pathways of energy systems. However, the low open-source availability of the frameworks in practice often leads to low interpretability and transparency of energy modeling system configurations. The configuration of an energy model entails how its system components, such as power plants, storage systems and grids operate, and which parameters are used to define them. In order to understand the impact of different model configurations and working principles on the model output, a thorough comparison between various modeling frameworks is necessary. This work thereby consists of a comparison of five open-source energy system modeling frameworks (OS-ESMFs) oemof , GENeSYS-MOD , Balmorel , urbs and GENESYS-2 on the mathematical level and spotlights selected methodological differences in renewable energy system modeling. The comparison shows diversity in the complexity of selected system components and helps to define the best use-cases and scales of application for each framework. Impacts of modeled features on the results were demonstrated by implementing two harmonized scenarios depicting the German electricity system using each framework. While similar model results were obtained for both scenarios, some differences were present, especially in the long-term expansion planning model. Some of those differences could be traced back to the identified modeling differences. • Selected mathematical features of five open energy modeling frameworks are compared. • A common terminology to facilitate direct mathematical comparison is developed. • Key feature differences are presented to increase transparency of modeling activities. • Influences of the modeling differences on the model results is demonstrated. • The frameworks show various differences, but remain flexible in specific application. [ABSTRACT FROM AUTHOR]